Re: A 3-Step Audit Protocol w/ 99%confidence

From: Kathy Dopp <kathy_dot_dopp_at_gmail_dot_com>
Date: Fri Jan 26 2007 - 23:45:05 CST

On 1/26/07, Charlie Strauss <cems@earthlink.net> wrote:
>
>
> >>
> >> I agree with arthur, this fudge factor is the achilles heel of the
> >> recount strategy. The good news however is two fold.
> >

>
> Kathy I think you need to re-read what I wrote. It clearly supports
> the auditing scheme you have propose, so obviously you misunderstood
> what I was saying.

Hi Charlie,

Thank you for continuing the conversation.

Did I misunderstand that you said that there is a "fudge factor" in
the audit methodology that is the achilles heel of the recount
strategy? Or did someone else say that and I was confused?

>
> I suspect you are somehow thinking the term fudge factor is a
> pejorative and that I am therefore opposing the auditing approach.
>
> I am not as your first question seems to think, suggesting
> eliminating the fudge factor. I'm for keeping it.

Well, it is impossible to eliminate the max vote miscount assumption
to have any logical consistency in the calculations. It cannot be
characterized as a fudge factor in the traditional use of that word.

>
> The point I was making is that there is a fudge factor because there
> is no agreement yet on what level woul dbe considered "NOT
> IMMEDIATELY NOTICEABLY SUSPICIOUS". If you re-read this you will
> see I was pointing to some rather alarmingly high rates in Sarasota
> FL that the judge did not consider suspcious enough.

That is true Charlie. I'm sorry if I didn't recognize your points.
However it is virtually irrelevant what value is picked for the
assumed maximum vote shift as long as the same value is used to
calculate what is suspicious and all suspicious vote counts are thrown
into the audit sample along with the randomly selected ones, as I've
mentioned in some of my footnotes and earlier emails on this subject.

>
> I thus took a moment to point out that raising this fudge factor
> towards higher and higher levels to the point where any judge would
> likely be moved to suspcision is going to lead to prohibitively high
> sampling rates.

Well, at most it can be moved up to 49.9% because beyond that is
physically impossible, and at that level (of 49.9%) is beyond
reasonable.

>
> Thus rather than throw the whole schema out, which is otherwise an
> elegantly simple scheme, because we can't all agree on where to set

Huh. It is not possible to "throw that whole schema out" if you want
to have any logical consistency in the design of election audits.
Please read this brief 3.5 page paper or my prior email again:

http://electionarchive.org/ucvAnalysis/US/paper-audits/FourTierAudit/TieredElectionAudits.pdf

> the fudge factor, I suggested we could essentially eliminate

Charlie, again it is a gross mischaracterization to call it a "fudge factor".

Again it is not important where one sets the max vote shift assumption
as far as the efficacy of the audit to detect outcome-altering vote
miscount, and the value you select would only effect the efficiency of
the audit procedures.

The value you select only effects how many vote counts are thrown into
the audit sample due to being suspicious versus due to being randomly
selected. Whatever value you select, you'll still have whatever
desired probability you want for detecting outcome-altering vote
miscount for the specific race.

I've used 20% to be ultra conservative in the design of my tiered
audits, since a 40% margin swing is huge and would be highly
suspicious, but used 15% in my calculated audits since that is the
high side of where the Brennan Center set the value in their
appendices.

> issue by having some stakeholder designated recount challenges
> (jargon: TAR) on top of the random selection of recounts for
> statistics. This will satisfy people better because when confronted
> with a judge who still does not think that a 25% apparent vote shift

A 25% vote shift from polls or whatever would mean a 50% margin shift
which is HUGE, but the point is that if the audit assumes that all
vote counts with more than a 15% partisan vote shift (30% partisan
margin shift) are suspicious then you MUST throw all vote counts with
more than that amount of partisan margin shift (of 30%) automatically
into the manual audit sample. It is part of the process of auditing
or the audit procedures are not logically coherent with the
assumptions that the audits are based on.

However research and development still needs doing to determine the
best ways to calculate that, as I mentioned Ron Rivest asked me how I
would suggest calculating it, but I am not planning to do any more
unpaid work on elections. I've had it with being in the center of a
circular firing squad of election integrity activists who either
wrongly steal credit for my work or attack and malign it wrongly
because they only want hand counted paper ballots without machines or
audits. I'm trying to remove myself soon from the madhouse unless
someone wants to fund me to finish the work.

Kathy
_______________________________________________
OVC-discuss mailing list
OVC-discuss@listman.sonic.net
http://lists.sonic.net/mailman/listinfo/ovc-discuss
==================================================================
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
==================================================================
Received on Tue Jan 1 14:12:50 2008

This archive was generated by hypermail 2.1.8 : Tue Jan 01 2008 - 14:12:51 CST